The Week AI Got Cheaper: What It Means for Your Business
Direct answer
The week of September 22, 2026, OpenAI launched GPT-6 Sol and GPT-6 Luna at prices roughly 50% below the previous generation, and Anthropic introduced Claude Opus 5.5 the same day at about 40% lower operating cost. In parallel, the Financial Times reported large enterprises migrating to open AI models to control costs. For your business the message is one: top-tier AI automation keeps getting cheaper, and the question now is which process you automate first.
What happened this week
- September 22: OpenAI launched GPT-6 Sol ($2 per million input tokens / $10 output) and GPT-6 Luna ($0.10 / $0.50), roughly 50% below GPT-5.6 promotional prices. Sol targets complex work and agents; Luna targets high-volume routine tasks. (TechRepublic, Crowdfund Insider, aiStockWire)
- Same day: Anthropic introduced Claude Opus 5.5: $4/$20 per million (20% below Opus 5) and about 40% lower operating cost on typical workloads. (aiStockWire) We already published the full analysis of this launch.
Two labs, the same day, the same direction: frontier AI is getting cheaper.
The other half of the story: enterprises move to open models
The Financial Times reported on September 27 that mentions of open models in earnings calls rose 6x in August and September versus the prior year (AlphaSense data), via PYMNTS and TradingView:
- AT&T runs about 40% of its AI workloads on open models and plans to reach 70% within a year.
- PNC Financial, C.H. Robinson, and Siemens are already discussing their uses publicly.
- Tinder's CTO said its AI spend went from $1M a year in January to $10M in July, and it's already routing routine queries to open models.
And the honest nuance: on Vercel's AI gateway, open models handled 56% of tokens in August but only 14% of spend; Anthropic captured 64% of total spend (via martincid.com). Translation for your business: cheap wins bulk work (summaries, support, follow-up); premium still gets paid where a mistake is expensive.
What it means for your business: 3 lessons
1. Cost stops being the excuse. If "AI is expensive" was an argument last year, this week buried it. What holds most businesses back is no longer price: it's not knowing where to start. Read what implementing AI costs in 2026.
2. The process matters, not the model. A well-designed system —a real AI agent— automatically gets cheaper and more capable with every launch, without rebuilding anything.
3. Start with volume. Customer support, follow-up, repetitive answers: that's where every cent saved per token multiplies across thousands of interactions.
Frequently asked questions
Should I switch models every time prices drop?
No. If your system is well designed, every price drop benefits you on its own. Switching for the sake of it breaks what works.
Are open models right for my business?
For high-volume routine tasks, yes. A serious provider picks the right model per task; you shouldn't have to think about it.
Will AI costs keep falling?
The trend of recent years is clear and accelerating. Every month you wait, automating gets cheaper for you — and for your competition.
What do I automate first at these prices?
What repeats every week and touches money: answering, qualifying, following up, scheduling.
At inicIA we turn artificial intelligence into business results: automations and agents that serve, qualify, and book while your team focuses on closing. If you want to see which process in your business gets automated first, let's talk on WhatsApp.
Chat on WhatsApp